A control method of a voltage equalization circuit of a low-voltage treatment device at the end of a low-voltage distribution network

Abstract

The application discloses a control method for a voltage equalization circuit of a low-voltage treatment device at the end of a low-voltage distribution network, and relates to the field of power distribution equipment.The control method comprises the following steps: real-time acquisition of multiple parameters; dynamic topology pre-judgment; voltage trend prediction results; selection of a compensation mode; hybrid energy storage collaborative control; energy storage SOC state judgment; execution of corresponding operations; multi-target compensation execution; and closed-loop feedback optimization.The control method for the voltage equalization circuit of the low-voltage treatment device at the end of the low-voltage distribution network realizes 100ms-level voltage trend prediction through an LSTM neural network, constructs a power grid digital twin through a dynamic link prediction algorithm, has high prediction accuracy, and realizes dynamic distribution of charging and discharging power through the complementation of supercapacitors and lithium batteries in power type and energy type.The NSGA-II genetic algorithm synchronously optimizes three targets of voltage compensation, harmonic suppression and energy storage protection, improves the harmonic suppression rate, and shortens the voltage sag recovery time.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment technology, specifically to a control method for the equalization circuit of a low-voltage management device at the end of a low-voltage power distribution network. Background Technology

[0002] At the end of low-voltage distribution networks, voltage sags and harmonic pollution are becoming increasingly prominent due to factors such as increased line impedance, severe load fluctuations, and the integration of new energy sources. Statistics show that the voltage qualification rate at the end of rural power grids in my country is less than 92%, and annual losses from industrial equipment downtime caused by voltage sags exceed 5 billion yuan. Traditional solutions have the following shortcomings:

[0003] Response lag: Reactive power compensation devices based on mechanical switches (such as SVC) have a response speed >50ms and cannot cope with rapid voltage fluctuations;

[0004] Inefficient energy storage management: individual energy storage units lack intelligent collaborative control, and lithium battery cycle life is less than 800 cycles;

[0005] Poor adaptability: unable to dynamically identify changes in power grid topology, and the compensation strategy is not targeted enough;

[0006] Limitations of harmonic mitigation: APF devices and reactive power compensation equipment operate independently, lacking multi-objective collaborative optimization. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a control method for the voltage equalization circuit of a low-voltage governance device at the end of a low-voltage distribution network. This method solves the problems of low voltage qualification rate at the end of rural power grids in my country and the limitations of traditional governance schemes, such as slow response, crude energy storage management, poor adaptability, and limited harmonic governance.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a control method for the equalization circuit of a low-voltage management device at the end of a low-voltage distribution network, comprising the following steps:

[0009] Step 1: Real-time acquisition of multiple parameters: The intelligent detection module acquires voltage, current, and harmonic parameters in real time through sensors;

[0010] Step 2: Dynamic Topology Prediction: The multi-modal control module predicts the power grid topology based on the collected parameters;

[0011] Step 3: Voltage Trend Prediction Results: The intelligent detection module analyzes the voltage trend and predicts whether the voltage will decrease or increase.

[0012] Step 4: Select the compensation mode: Based on the voltage trend prediction results, select either the supercapacitor priority compensation mode or the lithium battery excess energy absorption mode.

[0013] Step 5: Hybrid Energy Storage Coordinated Control: The multi-modal control module coordinates the hybrid energy storage execution unit to control the charging and discharging of the supercapacitor and lithium battery;

[0014] Step 6: Determine the SOC status of energy storage: Check the SOC status of the supercapacitor and lithium battery;

[0015] Step 7: Perform the corresponding operation: Based on the SOC status, activate lithium battery charging protection, perform multi-target compensation, or stop lithium battery charging;

[0016] Step 8: Multi-target compensation execution: The dynamic voltage equalization compensation circuit performs voltage compensation and harmonic suppression;

[0017] Step 9: Closed-loop feedback optimization: The system performs closed-loop feedback optimization based on the compensation effect.

[0018] Preferably, in the dynamic topology prediction step, the collected parameters are used to predict the power grid topology structure, and a dynamic link prediction algorithm is adopted to integrate network topology and attribute information to predict changes in the power grid structure, providing a basis for the selection of compensation mode.

[0019] Preferably, in the voltage trend prediction step, historical voltage data is analyzed using the LSTM algorithm to capture long-term dependencies, predict voltage rise and fall trends, and determine whether to choose supercapacitor compensation or lithium battery energy absorption mode.

[0020] Preferably, in the step of selecting the compensation mode, the voltage trend predicted by the LSTM network in the next 100ms is input for decision-making, and the decision rule is as follows:

[0021] If the predicted voltage is lower than the threshold, it can be set to 90% of the nominal voltage, which will activate the supercapacitor bank and use its high power density characteristics to quickly compensate.

[0022] If the predicted voltage rise exceeds the threshold, it can be set to 110% of the nominal voltage, in which case the lithium battery pack will be switched to absorb the excess energy to prevent overvoltage.

[0023] Preferably, the hybrid energy storage collaborative control adopts a model predictive control algorithm to optimize the charging and discharging power of the supercapacitor and lithium battery in real time, balancing power and energy demand.

[0024] Preferably, the energy storage SOC determination is performed by estimating the battery SOC using the ampere-hour integration method and the Kalman filter method, combined with the open-circuit voltage method for calibration.

[0025] Preferably, in the execution of the corresponding operation steps, the real-time SOC values ​​of the supercapacitor and lithium battery are input for decision-making, and the decision-making rule is as follows:

[0026] SOC<30%: The lithium battery enters charging protection mode, allowing only the supercapacitor to participate in instantaneous compensation;

[0027] 30%≤SOC≤80%: Dual energy storage works in tandem, with the supercapacitor handling high-frequency fluctuations and the lithium battery providing continuous support;

[0028] SOC>80%: The lithium battery stops charging and is used only as a backup power source to avoid the risk of overcharging.

[0029] Preferably, in the multi-target compensation execution step, the dynamic voltage equalization compensation circuit performs voltage compensation and harmonic suppression by adjusting the inductor current mode, thereby improving the voltage quality at the end of the power grid and suppressing harmonic interference.

[0030] Preferably, the closed-loop feedback optimization step uses the NSGA-II genetic algorithm to solve the multi-objective model and combines it with the analytic hierarchy process (AHP) to determine the optimal solution.

[0031] This invention discloses a control method for the voltage equalization circuit of a low-voltage management device at the end of a low-voltage distribution network, which has the following beneficial effects:

[0032] The control method of the voltage equalization circuit of the low-voltage governance device at the end of the low-voltage distribution network uses an LSTM neural network to achieve 100ms-level voltage trend prediction, a dynamic link prediction algorithm to construct a digital twin of the power grid with high prediction accuracy, supercapacitors and lithium batteries complement each other in terms of power type and energy type, and model predictive control algorithm to achieve dynamic allocation of charging and discharging power, extending the life of lithium batteries. The NSGA-II genetic algorithm simultaneously optimizes the three objectives of voltage compensation, harmonic suppression, and energy storage protection, improving the harmonic suppression rate and shortening the voltage sag recovery time. Through three core innovations—dynamic prediction, hybrid energy storage synergy, and multi-objective optimization—the voltage governance efficiency at the end of the low-voltage distribution network is significantly improved.

[0033] The control method of the voltage equalization circuit of the low-voltage management device at the end of the low-voltage distribution network involves an intelligent detection module that collects three-phase voltage / current through Hall sensors and Rogowski coils. An FFT algorithm simultaneously extracts harmonic components below the 50th order. After anti-aliasing filtering and 24-bit ADC conversion, the data is transmitted to the multi-mode control module via a dual-redundant CAN bus. The multi-mode control module performs dual prediction, using an LSTM network to analyze historical data within a 10-second time window to predict the voltage trend for the next 100ms. A dynamic link prediction algorithm is used to fuse topology data to predict the grid's operating status. Based on the prediction results, a compensation mode is selected: when the voltage is <90%UN, the supercapacitor bank is activated with a maximum discharge power of 500kW; when the voltage is >110%UN, the lithium battery bank is switched to absorb energy with a power absorption of 300kW. The hybrid energy storage execution unit receives PWM control signals. The supercapacitor injects compensation current through a three-level ANPC inverter, and the lithium battery adjusts its charging and discharging power through a bidirectional DC-DC converter. The dynamic voltage equalization compensation circuit simultaneously executes APF mode to compensate for voltage sags and SVG mode to suppress harmonics.

[0034] The control method for the voltage equalization circuit of the low-voltage management device at the end of the low-voltage distribution network consists of an intelligent detection module composed of a high-precision sensor array and a signal processing unit. The high-precision sensor array includes voltage and current sensors. The voltage sensor uses fiber optic sensing technology with a measurement accuracy of ±0.2% and a response time of <1μs. The current sensor is based on the Rogowski coil principle, with a range of 0-1000A and a linearity error of <0.5%. The signal processing unit uses a harmonic analysis unit with a built-in FFT algorithm, which can extract harmonic components below the 50th order in real time. It also uses a dual DSP architecture to handle real-time data acquisition and complex algorithm calculations, respectively. The voltage trend prediction algorithm is based on an LSTM network, with a historical data window of 10 cycles, predicting the voltage change trend in the next 100ms with a mean square error controlled within 0.005. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the overall process of the present invention;

[0037] Figure 2 This is a schematic diagram of the overall architecture of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] This application provides a control method for the voltage equalization circuit of a low-voltage governance device at the end of a low-voltage distribution network. This method addresses the problems of low voltage qualification rate at the end of rural power grids in my country, and the limitations of traditional governance schemes such as slow response, inefficient energy storage management, poor adaptability, and limited harmonic control. It achieves 100ms-level voltage trend prediction using an LSTM neural network, constructs a digital twin of the power grid with high prediction accuracy using a dynamic link prediction algorithm, complements the power type and energy type of supercapacitors and lithium batteries, and uses a model predictive control algorithm to achieve dynamic allocation of charging and discharging power, extending lithium battery life. The NSGA-II genetic algorithm simultaneously optimizes the three objectives of voltage compensation, harmonic suppression, and energy storage protection, improving harmonic suppression rate and shortening voltage sag recovery time.

[0040] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0041] This invention discloses a control method for the voltage equalization circuit of a low-voltage management device at the end of a low-voltage distribution network.

[0042] According to the appendix Figure 1-2 As shown, it includes the following steps:

[0043] Step 1: Real-time acquisition of multiple parameters: The intelligent detection module acquires voltage, current, and harmonic parameters in real time through sensors;

[0044] Step 2: Dynamic Topology Prediction: The multi-modal control module predicts the power grid topology based on the collected parameters;

[0045] Step 3: Voltage Trend Prediction Results: The intelligent detection module analyzes the voltage trend and predicts whether the voltage will decrease or increase.

[0046] Step 4: Select the compensation mode: Based on the voltage trend prediction results, select either the supercapacitor priority compensation mode or the lithium battery excess energy absorption mode.

[0047] Step 5: Hybrid Energy Storage Coordinated Control: The multi-modal control module coordinates the hybrid energy storage execution unit to control the charging and discharging of the supercapacitor and lithium battery;

[0048] Step 6: Determine the SOC status of energy storage: Check the SOC status of the supercapacitor and lithium battery;

[0049] Step 7: Perform the corresponding operation: Based on the SOC status, activate lithium battery charging protection, perform multi-target compensation, or stop lithium battery charging;

[0050] Step 8: Multi-target compensation execution: The dynamic voltage equalization compensation circuit performs voltage compensation and harmonic suppression;

[0051] Step 9: Closed-loop feedback optimization: The system performs closed-loop feedback optimization based on the compensation effect.

[0052] The LSTM neural network achieves 100ms-level voltage trend prediction, the dynamic link prediction algorithm constructs a digital twin of the power grid with high prediction accuracy, the supercapacitor and lithium battery are power-type and energy-type complementary, the model predictive control algorithm realizes dynamic allocation of charging and discharging power, and the lithium battery life is extended. The NSGA-II genetic algorithm simultaneously optimizes the three objectives of voltage compensation, harmonic suppression and energy storage protection, improves the harmonic suppression rate and shortens the voltage sag recovery time.

[0053] The intelligent detection module consists of a high-precision sensor array and a signal processing unit. The high-precision sensor array includes a voltage sensor and a current sensor. The voltage sensor uses fiber optic sensing technology with a measurement accuracy of ±0.2% and a response time of <1μs. The current sensor is based on the Rogowski coil principle, with a range of 0-1000A and a linearity error of <0.5%. The signal processing unit uses a harmonic analysis unit with a built-in FFT algorithm, which can extract harmonic components below the 50th order in real time. It also uses a dual DSP architecture to handle real-time data acquisition and complex algorithm calculations, respectively. The voltage trend prediction algorithm is based on an LSTM network, with a historical data window of 10 cycles, predicting the voltage change trend in the next 100ms with a mean square error controlled within 0.005.

[0054] The multimodal control module enables grid state prediction and energy storage coordinated control, constructs a grid topology model, and includes node attributes such as line impedance, load type, and distributed power generation output prediction. Based on the GraphSAGE framework, it adopts a dynamic link prediction algorithm to update the node connection probability matrix in real time, achieving a topology change prediction accuracy of >95%.

[0055] The hybrid energy storage execution unit adopts a hierarchical energy storage architecture, including a supercapacitor group and a lithium battery group. The supercapacitor group consists of 4 series 16.5V / 300F modules with a total capacity of 48V / 1200F. The lithium battery group consists of lithium iron phosphate cells with a capacity of 48V / 100Ah. It is equipped with a bidirectional DC-DC converter, and the SOC status monitoring adopts a fusion algorithm, specifically ampere-hour integration + Kalman filtering + open-circuit voltage method.

[0056] The dynamic voltage equalization compensation circuit adopts an improved active power filter structure. Its main circuit is a three-phase four-wire topology, with each phase equipped with an LCL filter. The inductance value is 1mH and the capacitance value is 20μF. The inductor current adjustment adopts direct current control, and the harmonic suppression ratio is >40dB. The closed-loop feedback optimization adopts the NSGA-II algorithm. The optimization objectives include minimizing voltage deviation and minimizing loss. The weight of minimizing voltage deviation is 0.6, and the weight of minimizing loss is 0.4. The Pareto front solution set selects the optimal compromise solution through the analytic hierarchy process.

[0057] The intelligent detection module collects three-phase voltage / current through Hall sensors, Rogowski coils, etc., and the FFT algorithm simultaneously extracts harmonic components below the 50th order. After the data is filtered by anti-aliasing and converted by a 24-bit ADC, it is transmitted to the multi-mode control module through a dual-redundant CAN bus.

[0058] The multimodal control module performs dual predictions. It uses an LSTM network to analyze historical data with a time window of 10 seconds to predict the voltage trend in the next 100ms. It uses a dynamic link prediction algorithm to fuse topology data to predict the grid operation status. Based on the prediction results, it selects a compensation mode. When the voltage is <90%UN, it activates the supercapacitor bank with a maximum discharge power of 500kW. When the voltage is >110%UN, it switches to the lithium battery bank to absorb energy with an absorption power of 300kW.

[0059] The hybrid energy storage execution unit receives PWM control signals, the supercapacitor injects compensation current through a three-level ANPC inverter, the lithium battery adjusts the charging and discharging power through a bidirectional DC-DC converter, the dynamic voltage equalization compensation circuit synchronously executes APF mode to compensate for voltage sags, and SVG mode to suppress harmonics.

[0060] Furthermore, in the dynamic topology prediction step, the collected parameters are used to predict the power grid topology structure. A dynamic link prediction algorithm is adopted to integrate network topology and attribute information to predict changes in the power grid structure, providing a basis for the selection of compensation mode.

[0061] Furthermore, in the voltage trend prediction step, historical voltage data is analyzed using the LSTM algorithm to capture long-term dependencies, predict voltage rise and fall trends, and determine whether to choose supercapacitor compensation or lithium battery energy absorption mode.

[0062] Specifically disclosed, in the step of selecting the compensation mode, the voltage trend predicted by the LSTM network over the next 100ms is input for decision-making, and the decision rule is as follows:

[0063] If the predicted voltage is lower than the threshold, it can be set to 90% of the nominal voltage, which will activate the supercapacitor bank and use its high power density characteristics to quickly compensate.

[0064] If the predicted voltage rise exceeds the threshold, it can be set to 110% of the nominal voltage, in which case the lithium battery pack will be switched to absorb the excess energy to prevent overvoltage.

[0065] Furthermore, the hybrid energy storage collaborative control adopts a model predictive control algorithm to optimize the charging and discharging power of the supercapacitor and lithium battery in real time, balancing power and energy demand.

[0066] Furthermore, the energy storage SOC determination is achieved by estimating the battery SOC using the ampere-hour integration method and the Kalman filter method, combined with the open-circuit voltage method for calibration.

[0067] Specifically disclosed, the real-time SOC values ​​of the supercapacitor and lithium battery are input for decision-making during the execution of the corresponding operation steps, and the decision-making rule is as follows:

[0068] SOC<30%: The lithium battery enters charging protection mode, allowing only the supercapacitor to participate in instantaneous compensation;

[0069] 30%≤SOC≤80%: Dual energy storage works in tandem, with the supercapacitor handling high-frequency fluctuations and the lithium battery providing continuous support;

[0070] SOC>80%: The lithium battery stops charging and is used only as a backup power source to avoid the risk of overcharging.

[0071] Furthermore, in the multi-target compensation execution step, the dynamic voltage equalization compensation circuit performs voltage compensation and harmonic suppression by adjusting the inductor current mode, thereby improving the voltage quality at the end of the power grid and suppressing harmonic interference.

[0072] Furthermore, in the closed-loop feedback optimization step, the NSGA-II genetic algorithm is used to solve the multi-objective model, and the optimal solution is determined by combining it with the analytic hierarchy process.

[0073] The LSTM neural network achieves 100ms-level voltage trend prediction, the dynamic link prediction algorithm constructs a digital twin of the power grid with high prediction accuracy, the supercapacitor and lithium battery are power-type and energy-type complementary, the model predictive control algorithm realizes dynamic allocation of charging and discharging power, and the lithium battery life is extended. The NSGA-II genetic algorithm simultaneously optimizes the three objectives of voltage compensation, harmonic suppression and energy storage protection, improves the harmonic suppression rate and shortens the voltage sag recovery time.

[0074] The intelligent detection module collects three-phase voltage / current through Hall sensors, Rogowski coils, etc., and the FFT algorithm simultaneously extracts harmonic components below the 50th order. After the data is filtered by anti-aliasing and converted by a 24-bit ADC, it is transmitted to the multi-mode control module through a dual-redundant CAN bus.

[0075] The multimodal control module performs dual predictions. It uses an LSTM network to analyze historical data with a time window of 10 seconds to predict the voltage trend in the next 100ms. It uses a dynamic link prediction algorithm to fuse topology data to predict the grid operation status. Based on the prediction results, it selects a compensation mode. When the voltage is <90%UN, it activates the supercapacitor bank with a maximum discharge power of 500kW. When the voltage is >110%UN, it switches to the lithium battery bank to absorb energy with an absorption power of 300kW.

[0076] The hybrid energy storage execution unit receives PWM control signals, the supercapacitor injects compensation current through a three-level ANPC inverter, the lithium battery adjusts the charging and discharging power through a bidirectional DC-DC converter, the dynamic voltage equalization compensation circuit synchronously executes APF mode to compensate for voltage sags, and SVG mode to suppress harmonics.

[0077] The intelligent detection module consists of a high-precision sensor array and a signal processing unit. The high-precision sensor array includes a voltage sensor and a current sensor. The voltage sensor uses fiber optic sensing technology with a measurement accuracy of ±0.2% and a response time of <1μs. The current sensor is based on the Rogowski coil principle, with a range of 0-1000A and a linearity error of <0.5%. The signal processing unit uses a harmonic analysis unit with a built-in FFT algorithm, which can extract harmonic components below the 50th order in real time. It also uses a dual DSP architecture to handle real-time data acquisition and complex algorithm calculations, respectively. The voltage trend prediction algorithm is based on an LSTM network, with a historical data window of 10 cycles, predicting the voltage change trend in the next 100ms with a mean square error controlled within 0.005.

[0078] The multimodal control module enables grid state prediction and energy storage coordinated control, constructs a grid topology model, and includes node attributes such as line impedance, load type, and distributed power generation output prediction. Based on the GraphSAGE framework, it adopts a dynamic link prediction algorithm to update the node connection probability matrix in real time, achieving a topology change prediction accuracy of >95%.

[0079] The hybrid energy storage execution unit adopts a hierarchical energy storage architecture, including a supercapacitor group and a lithium battery group. The supercapacitor group consists of 4 series 16.5V / 300F modules with a total capacity of 48V / 1200F. The lithium battery group consists of lithium iron phosphate cells with a capacity of 48V / 100Ah. It is equipped with a bidirectional DC-DC converter, and the SOC status monitoring adopts a fusion algorithm, specifically ampere-hour integration + Kalman filtering + open-circuit voltage method.

[0080] The dynamic voltage equalization compensation circuit adopts an improved active power filter structure. Its main circuit is a three-phase four-wire topology, with each phase equipped with an LCL filter. The inductance value is 1mH and the capacitance value is 20μF. The inductor current adjustment adopts direct current control, and the harmonic suppression ratio is >40dB. The closed-loop feedback optimization adopts the NSGA-II algorithm. The optimization objectives include minimizing voltage deviation and minimizing loss. The weight of minimizing voltage deviation is 0.6, and the weight of minimizing loss is 0.4. The Pareto front solution set selects the optimal compromise solution through the analytic hierarchy process.

[0081] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A control method for the voltage equalization circuit of a low-voltage management device at the end of a low-voltage distribution network, characterized in that, Includes the following steps: Step 1: Real-time acquisition of multiple parameters: The intelligent detection module acquires voltage, current, and harmonic parameters in real time through sensors; Step 2, Dynamic Topology Prediction: The multi-modal control module predicts the power grid topology based on the collected parameters. It uses a dynamic link prediction algorithm to integrate three types of information: line impedance, load type, and distributed power generation output prediction values ​​to construct a digital twin of the power grid and predict changes in the power grid structure. Step 3: Voltage Trend Prediction Results: The intelligent detection module analyzes historical voltage data through an LSTM neural network, uses a 10-cycle historical data window to capture long-term dependencies, and achieves 100ms-level voltage trend prediction to determine whether the voltage is decreasing or increasing. Step 4: Select the compensation mode: Based on the voltage trend prediction results, if the predicted voltage is lower than the threshold, the supercapacitor group is activated to quickly compensate using its high power density characteristics; if the predicted voltage rises above the threshold, the lithium battery group is switched to absorb the excess energy. Step 5, Hybrid Energy Storage Coordinated Control: The multi-modal control module adopts a model predictive control algorithm to coordinate the dynamic allocation of charging and discharging power of the supercapacitor and lithium battery controlled by the hybrid energy storage execution unit, balancing power and energy demand; Step 6: Energy Storage SOC Status Determination: Estimate the battery SOC using the ampere-hour integration method and Kalman filtering method, and calibrate using the open-circuit voltage method to check the SOC status of the supercapacitor and lithium battery; Step 7: Perform the corresponding operation: Based on the SOC status, activate lithium battery charging protection, perform multi-target compensation, or stop lithium battery charging; Step 8: Multi-target compensation execution: The dynamic voltage equalization compensation circuit performs voltage compensation and harmonic suppression by adjusting the inductor current mode, which improves the harmonic suppression rate and shortens the voltage sag recovery time. Step 9: Closed-loop feedback optimization: The NSGA-II genetic algorithm is used to simultaneously optimize the three objectives of voltage compensation, harmonic suppression, and energy storage protection. The optimal solution is determined by combining the analytic hierarchy process (AHP). The weight of minimizing voltage deviation is 0.6, and the weight of minimizing loss is 0.

4. The system performs closed-loop feedback optimization based on the compensation effect.

2. The control method for the voltage equalization circuit of a low-voltage distribution network end-of-line low-voltage management device according to claim 1, characterized in that, The dynamic topology prediction step utilizes collected parameters to predict the power grid topology and forecast changes in the power grid structure, providing a basis for selecting the compensation mode.

3. The control method for the voltage equalization circuit of a low-voltage distribution network end-of-line low-voltage management device according to claim 1, characterized in that, In the voltage trend prediction step, the LSTM algorithm is used to capture long-term dependencies, predict voltage rise and fall trends, and decide whether to choose supercapacitor compensation or lithium battery energy absorption mode.

4. The control method for the voltage equalization circuit of a low-voltage distribution network end-of-line low-voltage management device according to claim 1, characterized in that, The decision-making process involves inputting the real-time SOC values ​​of the supercapacitor and lithium battery during the execution of the corresponding operation steps, and the decision-making rule is as follows: SOC<30%: The lithium battery enters charging protection mode, allowing only the supercapacitor to participate in instantaneous compensation; 30%≤SOC≤80%: Dual energy storage works in tandem, with the supercapacitor handling high-frequency fluctuations and the lithium battery providing continuous support; SOC>80%: The lithium battery stops charging and is used only as a backup power source to avoid the risk of overcharging.

Images